Many researchers have illustrated the difficulties and needs that children and adults have with formal and informal reasoning (e.g., Kuhn, 1989, 1993; Perkins, Allen, & Hafner, 1983). The Theory of Explanatory Coherence (TEC) and its associated connectionist model, ECHO, offers an account of how people decide the plausibility of beliefs asserted in an explanation or argument (e.g., Ranney & Thagard, 1988; Thagard, 1989). We have found that ECHO usefully predicts how people evaluate hypotheses, evidence, and other propositions regarding various situations (Schank & Ranney, 1991 & 1992). Insights from these descriptive studies led us to develop a prescriptive ECHO-based "reasoner's workbench" computer program Convince Me and an associated scientific reasoning curriculum to help students structure, restructure, and assess their knowledge about often controversial situations (e.g., Schank & Ranney, 1993). Using Convince Me, students can (a) easily create, modify, load, and save arguments, (b) rate how strongly they believe each statement in the argument, and (c) run an ECHO simulation to see which statements their argument helped to support or reject (and which ones it left neutral) from ECHO's point of view. Disparities between students' own evaluations and ECHO's can help them pinpoint inconsistencies in their arguments. As a result, they may re-evaluate their beliefs, reformulate their arguments, or even adjust ECHO's numerical parameters to better model their way of thinking.

Does Convince Me significantly help students articulate and revise their theories? This dissertation discusses my (and others') prior work on modeling and aiding reasoning, then assesses the effectiveness of the Convince Me system. Also addressed are questions regarding the evidence/hypothesis distinction and effects of context on reasoning. Results suggest that although the distinguishing characteristics of data and theory are vague--even for experts who study scientific reasoning professionally--Convince Me lends a sophistication to novices' discriminative criteria across contexts, making their epistemic categorizations more expert-like both during, and after, its use. Further, more accurate and honest portrayals of these constructs as fuzzy and dependent on context may help students view science as a dynamic field that requires the continuous examination and revision of ideas, rather than the memorization of disconnected "facts."

Download the postscript version of the dissertation.

See the Convince Me home page to try out the software and see related papers.

Copyright © 1999 Patti Schank, All Rights Reserved.